Comparative analysis of machine learning algorithms for classification of services in LTE/5G networks based on QoS parameters

DOI: 10.31673/2412-9070.2026.017414

Abstract

The article considers the problem of automatic classification of service types in LTE/5G mobile networks based on key quality of service (QoS) parameters. The need for such classification arises from the diverse requirements of modern mobile services, including Ultra-Reliable Low-Latency Communications (URLLC), enhanced Mobile Broadband (eMBB), and massive Machine-Type Communications (mMTC), which differ significantly in terms of throughput, latency, jitter, packet loss, and traffic patterns. To conduct the study, a simulated dataset was generated to realistically reflect the statistical characteristics of these parameters, including forward and reverse traffic ratios, and incorporating noise distortions and random label errors to better approximate real network conditions. This approach ensures that the experimental evaluation accounts for uncertainties and variations typical of operational LTE/5G environments.
A comparative analysis was carried out using several widely known machine learning models, including logistic regression, linear and radial basis function (RBF) support vector machines, random forest, k-nearest neighbors, naive Bayesian classifier, and decision tree. Each model was trained on a portion of the dataset and tested on the remaining samples to evaluate classification accuracy. The results of this analysis are presented graphically, highlighting the accuracy ranges of each algorithm and indicating that ensemble and kernel-based methods, particularly random forest and RBF SVM, outperform other approaches under the given experimental conditions.Different machine learning algorithms demonstrate their effectiveness depending on the nature of the data, the level of noise and the nonlinearity of the dependencies between network parameters. This necessitates the need for a comparative analysis of classification models in order to select the most suitable solutions for practical application in LTE/5G networks.
The study demonstrates that different machine learning algorithms exhibit varying performance depending on the nonlinearity of data, the presence of noise, and the overlapping characteristics of service classes. Accurate classification of service types based on QoS parameters is critical for effective network management, optimal resource allocation, and ensuring consistent quality of experience for end users. This underscores the importance of performing a comparative evaluation of multiplemodels to identify the most suitable algorithms for practical deployment in real LTE/5G networks. The findings of this research contribute to the development of intelligent network monitoring and management systems, providing guidance for future work in automated service classification and optimization in heterogeneous mobile networks.

Keywords: machine learning; dataset; bandwidth; ML-model; network traffic analysis; classifycation algorithms; quality of service; network resource optimization; service type.